Journal of Environmental Science and Management 25-1: 20-33 (June 2022) ISSN 0119-1144
Ana Rosa A. Carmona1*
Victor B. Ella2
1 College of Education, University of Saint La Salle, Bacolod City
2 Land and Water Resources Division, College of Engineering and Technology, University of the Philippines-Los Baños, Laguna 4031
*Corresponding author:
[email protected] ABSTRACT
Catchment classification is one approach in natural resource management that is widely adopted in taking efficient steps towards implementing suitable soil and water conservation measures across a basin or region. Catchments have unique characteristics emerging from the heterogeneity and complexity of the systems and classifying them paves way to achieve order and simplicity. However, some constraints related to data availability could be a problem in a region where only few rivers are gauged and with only one type of climate data available. This study presents a way to decrease complexity by grouping these catchments based on their biophysical characteristics extracted from readily available datasets and using simple statistical approaches. Principal component analysis was first conducted to twenty-four biophysical variables which were reduced to eight factor components. A hierarchical clustering method was then performed to define the number of clusters and K-means clustering procedure was followed for the final grouping. Nine watershed clusters were formed with watershed size having the greatest contribution. Grouping catchments into clusters with similar biophysical characteristics does not only promote simplicity but also facilitates understanding of the nature of not only one watershed but also its relationship with other watersheds in a bigger landscape. The study also confirmed that spatially close watersheds exhibit similar characteristics.
Keywords: catchment classification, cluster analysis, biophysical characteristics, physical similarity, watershed management
INTRODUCTION
One of the existing challenges in the field of hydrology is coming up with a generally agreed catchment classification system (Wagener et al. 2007;
Ali et al. 2012; Salami and Bueher 2020) and framework (Sawicz et al. 2011; Sivakumar et al. 2014). Catchments are generally complex and dynamic environmental systems, with diverse components interacting with each other, where fluxes of energy, water and nutrients flow in and out of the landscape. Beven (2000) put forward the idea of “uniqueness of place”, which recognizes the distinct characteristics of a catchment with respect to difference in hydro-climatic condition, soil, topography, geology, vegetation, and human modification through various temporal and spatial scales. However, despite the overwhelming complexities and differences among catchments, Wagener et al.
(2007) believed that patterns and connections might be discernable, which could shed light to a way to put forth order among various watersheds. They further pointed out that a proposed catchment classification
JESAM
Exploring Homogeneity among Catchments for Efficient Province-wide Watershed Management in Negros Occidental, Philippines
should be a rigorous scientific inquiry into the causes of similarities and relationships among catchments.
The uniqueness of a watershed is brought about by the interaction of its various internal and external components in time and space. These characteristics define the uniqueness of the watershed (Beven 2000) and affect its hydrological processes and in turn affect the socio- ecological system embedded within the watershed (Chess and Gibson 2001; O’Neill 2005; Carillo et al. 2011; Mayer et al. 2014). A systematic classification of the catchments based on patterns of their physical structures is one easier way to find order and to better understand these processes amidst of heterogeneity and complexity in a region.
At present, catchments are classified using different descriptive terms including those based on land cover (forested, agricultural, urban, etc.), storage (groundwater dominated, surface water dominated), catchment response (fast, slow), and catchment size (basin, watershed, sub-
21 and most catchment studies are catchment-specific.
There are thousands of watersheds of various sizes in the Philippines – most of which are found within the 421 principal river basins. The reviewed records showed that more than 270 studies were conducted in 160 Philippine rivers from 1975 until 2014. Most of these studies focused on watershed management and planning (19%) and followed by topics related to hydrology (11.3%), climate change (10.6%), biodiversity (10.4%), and soil resources (8.8%). More than 50% of the watershed studies, clumped under the main topic of hydrology, were related to the use of hydrologic modeling in prediction of the effect of climate, land cover change, and management approach to run-off and sediment load. Other studies also focused on the relationship of soil properties, climate, and land cover with the hydrologic processes and with some hydrological assessments, stream flow and water yield quantification.
Most of these catchment hydrology studies in the Philippines focused more on a specific river basin or watershed with limited application to other watersheds especially for ungauged rivers or in data-poor context regions. Studies on comparison, regionalization, classification, or prioritization of more than two watersheds based on similar characteristics are limited in local literature. More so, in the Island of Negros where only very few watersheds had been studied and gauged. The adage, “what gets measured gets managed”, provides a good reason for the need to study, quantify and act on catchment patterns, processes and functions using the watershed as a unit of landscape study and for management. Hence, a regionalization study covering 33 watersheds found in the Island of Negros was conducted from 2016 to 2017.
Catchment classifications were generally based on the concept of relating climate and catchment characteristics to hydrologic behavior (McDonnell and Woods 2004; Carillo et al. 2011; Vasquez et al. 2019).
These could be simplified into three types of similarities, the physiographic, climatic and hydrologic similarities.
Based on reviewed studies, various climate, catchment and hydrologic characteristics and their combinations have been used in catchment classifications. There are those, which focused only on hydrologic components (Sawicz et al. 2011; Brown et al. 2014; Salami and Buehler 2020) or catchment physical structures (Bayrd 2006), those, which combined the physiographic and hydrological components (Breinlinger et al. 1996; Mazvimavi 2003; Carillo et al.
2011; Ali et al. 2012; Rasavi and Coulibaly 2013; Yi et al. 2018; van Hoek 2019; Hidayatullo et al. 2021) or physiographic and climate variable (Raux et al. 2011), watershed, mini-watershed, micro-watershed). These
classification schemes however are considered limited and not well-defined (Wagener et al. 2007). The way forward is to come up with metrics that define the similarity or dissimilarity among catchments. These metrics are based on understanding the relationship between the structural features of the catchments (geomorphology, pedology, geology, vegetation) and climate variability to define hydrological response characteristics (streamflow, groundwater, soil moisture) of the catchments (Carillo et al. 2011; Vasquez et al. 2019). Sivakumar et al. (2014) proposed a catchment classification framework, which also includes other components such as catchment processes, hydroclimatic processes, other natural and anthropogenic factors, and the interaction of surface water, soil moisture and groundwater components.
There were several documented attempts of catchment grouping in other parts of the world using different bases. Small basins in Switzerland were classified using 37 parameters of catchment characteristics relevant to discharge and consequently their similar hydrological behavior with the use of a two- way indicator species analysis (Breinlinger et al.
1996). In Iran, Choubin et al. (2017) used remote sensing indices to identify homogenous hydrological watersheds among 38 catchments. Bayrd (2006) also grouped 17 catchments found in the United States through Principal Component Analysis (PCA) based on their lithologic and morphologic characteristics. Mazvimavi (2003) classified 52 catchments in Zimbabwe using catchment characteristics and flow characteristics through Ward’s clustering technique. Raux et al. (2011) also classified 24 large drainage basins in the world using geomorphologic and climatic variables utilizing multivariate statistical analyses such as Cluster Analysis and PCA. Affinity Propagation clustering technique was performed on 18 sub-catchments in the Kingdom of Saudi Arabia (Hidayatulloh et al. 2021) and on 36 Scottish sites (Ali et al.
2012). In Chile, Vasquez et al. (2019) utilized a Bayesian clustering algorithm to classify 82 basins using physical, climatic, and hydrologic parameters. K-means algorithm was also used to classify catchments in upper Huai River Basin, China (Yi et al. 2018). These are just few of the many successfully conducted catchment classifications based on different sets of catchment characteristics and other external factors, covering different temporal and spatial scales, using different models and analysis tools, and for varying purposes. While there are many ways to classify catchments, most of the developed models have many parameters and require too much data (Sivakumar et al. 2014). Data availability is the main constraint in many areas in the Philippine archipelago
22
ranging from 30 to 2,100 km2 with outlets found in 25 coastal cities and municipalities of the province.
Datasets
The readily available datasets utilized for the characterization and quantitative analysis of the catchment were accessed from different government offices such as National Mapping Authority and Resources Authority (NAMRIA), Mines and Geosciences Bureau (MGB), Bureau of Soils and Water Management (BSWM) (Table 1). Thematic maps were digitized, geo-referenced and projected following the coordinate system, WGS 1984 UTM Zone 51N, using ArcGIS software. Some processed shapefiles and raster data were also accessed from the Department of Agriculture-Bureau of Agricultural Research (DA-BAR) and from the PhilGIS website.
Watershed Biophysical Characterization
Geomorphology. Quantitative analyses of the land surface of watersheds were carried out using vector and those, which combined the catchment physical
structures with climate and hydrological data (Sawicz et al. 2011; Sawicz 2013; Vasquez et al. 2019). As there is no globally-agreed, broad-scale catchment classification system, the choice of parameters used in catchment classification in many cases is assessed based on the purpose of the classification and the available data. Considering the limitation on available hydrologic data, where only 6 out 33 rivers are gauged, and there is only one meteorological station in the province, this study only considered the catchment biophysical parameters, which are readily available from government offices. Geomorphometric, land cover, lithologic, and soil variables representing the catchment physical structures were measured, analyzed, and utilized for catchment characterization and classification. The primary purpose of this catchment classification is to reduce the complexity of 33 catchments by grouping them into similar categories based on catchment structures for locations with minimal available data.
Moreover, classification of watersheds into clusters based on homogenous catchment characteristics would allow for a more area-specific approach to soil and water conservation. Catchment characterization and classification involve valuable information that could be utilized for island-wide planning and management of water and soil resources using watershed as a management unit. These are also preliminary steps toward flood and soil erosion susceptibility analysis, flood vulnerability analysis of population in flood prone areas of the watershed, catchment condition analysis, and catchment prioritization.
MATERIALS AND METHODS Study Area
The Island of Negros is the fourth largest island in the Philippine archipelago comprising of two provinces (Negros Occidental and Negros Oriental), one highly urbanized city (Bacolod City), 18 component cities, 38 municipalities, and a total of 1,219 barangays (Figure 1).
It is a volcanic island that is predominantly an agricultural region. More than 57% of its total land area is cultivated with sugarcane as its main crop, making it the country’s main producer of sugar (Guadalquier and Nicavera 2019).
Most of the cities and towns of Negros Occidental are found in coastal areas where 109 rivers empty into the bigger bodies of water. The study covered 33 major watersheds with river systems having stream orders of more than three. These watersheds have catchment areas
Figure 1. Location of major watersheds in the Province of Negros Occidental, Philippines.
23
data extracted from stream networks and topographic lines. Basic morphometric parameters were determined using measuring tools in GIS while more advanced morphometric parameters were determined following standard derived formulas (Table 2). Relief parameters were determined using values extracted from the digital elevation model (DEM). Fifteen geomorphometric parameters were chosen based on the Department of Environment and Natural Resources (DENR 2008) recommendation for watershed characterization and
those commonly found in reviewed literature. Five parameters were used to describe watershed size and shape (basin area, mean bifurcation ratio, elongation ratio, circularity ratio, and form ratio) and six parameters to describe watershed relief (relief ratio, basin mean elevation, hypsometric index, ruggedness number, and slope). Drainage characteristics of the watershed were described by drainage density, stream frequency, stream length ratio, length of overland flow, and stream order.
Table 1. Map types, scales and sources used for geomorphology, geology, soil, and land cover parameters..
Parameter Type Scale Source
Geomorphology Geology SoilLand Cover
Digital Elevation Model Topographic map
Geologic Map with accompanying report
Soil Map with accompanying report (1951, 1960) Land cover (2010)
30-m/90-m 1:50,000 1:250,000 1:200,000 (Shapefile)
ASTER/SRTM NAMRIA
BSWMMGB DA-BAR
Table 2. Methodology adopted for the computation of morphometric parameters.
Parameter Description Symbol/Formula Reference
1. Basin Area 2. Mean Bifurcation
Ratio
3. Elongation Ratio 4. Circularity Ratio 5. Form Factor Ratio 6. Relief Ratio 7. Basin Mean
Elevation 8. Hypsometric
Integral
9. Ruggedness Number
10. Slope
11. Drainage Density
12. Stream Frequency 13. Stream Length
Ratio
14. Length of Overland Flow
15. Stream Order
The area of bounded watershed that contributes to surface run-off
The average of bifurcation ratio of all orders. Bifurcation ratio is the ratio between the number of stream segments of one order and those of the next-higher order.
Ratio of diameter of a circle of the same area as the basin to the maximum basin length
Ratio of basin area to the area of circle having the same perimeter as the basin
Ratio of basin area to the square of basin length
Difference in elevation between the river’s source and the river’s mouth divided by the river’s total length The mean watershed altitude
Generally summarizes the basin’s relief; a general index of erosional development
Expresses the geometric characteristics of a drainage system; derived from the product of maximum basin relief and drainage density
Percentage of area with slope of more than 18%
Ratio of the total channel segment lengths cumulated for all orders within a basin to the basin area
Number of stream segments per unit area
Ratio of the mean of segments of a given order to the (cumulative) mean length of the next lower order stream in the same basin
Length of the run of the rainwater on the surface before it is localized into definite channels; roughly equals to half the reciprocal of drainage density
Based on Strahler’s method following the hierarchical rank
A
Rr=Br/Bl BE
Rn=Bh(Dd)
S
Fs=Nu/A
Lof=1/Ddx2
U
Al-Rawas &
Valeo 2010 Schumm1956
Schumm 1956 Miller 1953 Horton 1932 Schumm 1956 Strahler 1952 Strahler 1952 Strahler 1968
Chorley 1979 Horton 1945;
Strahler 1952;
Melton 1958 Horton 1932 Horton 1945
Horton 1945
Strahler 1964
24
Hypsometric index was also determined to provide additional information on the watershed relief characteristics following Brocklehurst and Whipple (2004) method. The study utilized the hypsometry extension software of ArcGIS (Davis 2010) to automatically generate the hypsometry curve. From a projected DEM clipped to the boundary of the watershed, float values were converted to integer format and were reclassified into 100 elevation bins using spatial analyst tools. The values were used in the calculation of the hypsometric integral (HI) which provides a general index of erosional development within the watershed.
Geology. A clipped shapefile for each watershed was prepared to determine the covered lithologic units and their corresponding extent in percentage. The description of lithologic units was based on the report, Geology and Mineral Resources of the Philippines (MGB 1981). The lithologic types which greatly affect the flow characteristics focused on the permeability of the lithologic units were chosen as geologic indicators.
Seventeen lithologic units found in thirty-three watersheds were reduced into three major classes, namely: igneous, sedimentary, and quaternary alluvium (Figure 2.A).
Soils. Proportions of soil types found within each watershed were identified based on the government map.
Thirty-four soil types were combined according to their soil textural class based on the percentage of sand, silt, and clay. Three groups of soils were formed: heavy, medium, and light soils.
Land Cover. Built-up, cultivated, and forest land cover types were chosen as they are the land cover types which have significant impact on the hydrologic behavior of the landscape. The area covered was expressed in percentage.
Watershed Classification
Classification was based on similarities of their biophysical characteristics geomorphometric (15), geologic (3), soil (3), and land cover (3) variables.
Factor analysis, particularly PCA, was conducted on the standardized computed values of the selected variables for each watershed to avoid multicollinearity among variables and for data reduction. The PCA also helped in finding a linear combination of variables that accounts for as much variation in the original variables as possible.
Using Eigenvalue of 1.0 and based on rotated component matrices, the number of factors for cluster analysis was reduced to resolve duplication of correlated variables.
The 24 variables reduced to eight components were
determined to be significant by comparison of the variance explained by each component. These were further used in watershed classification. Cluster analysis was used to determine groups of watersheds that are relatively homogeneous within themselves and heterogeneous among each other, based on their watershed biophysical characteristics.
Ward’s method, a hierarchical clustering method, was first performed to define the number of clusters based on the agglomeration schedule and the dendrogram Ward’s method, a hierarchical clustering method, was firstperformed to define the number of clusters based on the agglomeration schedule and the dendrogram produced. K-means clustering procedure was done for actual grouping of the watersheds. All statistical analyses were done using the SPSS statistical program, which also provided statistical tables showing additional information to better understand the nuances of clustering.
RESULTS AND DISCUSSIONS Cluster Formation
Factor analysis with varimax rotation reduced the twenty-four biophysical variables into eight components which explain 85.14% of the total variations of the dataset. The first component, describing the drainage characteristics of the watershed, explained the highest variation (24.76%). The second component, explained 17.99% variance, consists of variables describing the shape of the watershed. The third component (11.58%) has variables related to the relief characteristics of the watershed. The fourth component (8.38%) consists of area and stream order, which both provided an idea of the extent of the watershed size. The lithologic units, igneous and sedimentary, together with cultivated area make up the fifth component, which explained 7.71% of the variance. The sixth component (5.57%) combined the variables explaining the stream network of the watershed and the extent of the alluvial deposits in the catchment.
The seventh component (4.81%) is contributed by soil texture, particularly the clayey and loamy soils. The last component (4.34%) represents land cover characteristics of the catchment with a contribution of the lighter- textured soils (Table 3).
The resulting component scores applied in the hierarchical cluster analysis using Ward’s method produced a dendrogram which presents the distance at which clusters were joined and at what level any two clusters were joined (Figure 3). A cluster distance of not more than ten was set and nine clusters were
25
Figure 2. Drainage system (A), Geologic map (B), Soil map (C), and Land cover man (D) of the Negros Island.
26
initially identified. This was used as input cluster number in K-means cluster analysis.
The F ratio values described the differences for each factor among clusters (Table 5). Watershed size showed a higher contribution (F = 11.61) in the formation of clusters.
This is followed by land cover variables (F = 9.02),
variables describing the drainage characteristics (F = 8.58), watershed shape variables (F = 7.68), catchment geology (F = 6.97), and soil variables (5.08). Watershed relief and variables describing the stream characteristics indicate least contribution to the separation of clusters.
Cluster analysis also provided additional information about the general characteristics of the clusters formed from the final cluster centers table provided by SPSS.
The regression factor scores provided an idea on how well each component predicted the score for each of the analyses (Table 6). Cluster 1 has high scores from variables pertaining to watershed size and drainage characteristics. Cluster 2 is defined by medium and heavy soils. Watersheds in Cluster 3 are strongly influenced by their stream characteristics and proportion of alluvium.
Cluster 4 watersheds are distinct by their land cover characteristics and light-textured soil. Cluster 5 and Cluster 9 are grouped based on their relief characteristics with a bit of influence on drainage characteristics for Cluster 5 and watershed shape for Cluster 9. Cluster 6 watersheds are identified by their watershed size and a bit from their shape and stream characteristics. Cluster 7 is strongly influenced by their lithological characteristics and Cluster 8 has higher scores in watershed drainage and shape factors.
Watershed Clusters
Narrative descriptions of the inherent characteristics of each watershed cluster type are discussed below together with their spatial location in the region and their predicted level of susceptibility to flood and erosion (Figure 4).
Cluster 1 – Watersheds with a large catchment area and few, long stream networks. The large drainage area of Ilog-Hilabangan river basin of around 2,109 km2 makes it look like a veritable giant compared to more than half of the major watersheds with less than 100 km2 Figure 3. Classification tree of watersheds produced
from hierarchical cluster analysis.
Table 3. Component variables and variance explained by each component.
Component Variables Percent Variance General Characteristics
1 23 45 67 8Total
Length of overland flow, Drainage density, Stream frequency, Hypsometric integral
Elongation ratio, Form factor ratio, Circularity ratio
Relief ratio, Slope, Basin mean elevation, Ruggedness number Area, Stream order
Igneous, Cultivated, Sedimentary
Mean bifurcation ratio, Stream length ratio, Alluvium Medium soil, Heavy soil
Forested area, Built-up area, Light soil (a total of 24 variables)
24.76 17.99 11.58 8.387.71 5.574.81 85.144.34
Watershed drainage Watershed shape Watershed relief Watershed size
Geology Stream characteristics
Soil Land cover
27
catchment area. Its immense size and high stream order set it apart from other watersheds. Drainage variables such as length of overland flow, stream frequency, drainage density, as well as hypsometric integral are all related to area. Hence, these also contribute to the uniqueness of this catchment. Ilog-Hilabangan is one of the watersheds with low values in drainage density and stream frequency and high length of overland flow and stream order number. Its fewer number and longer length of streams result in relatively longer travel distance of runoff before it gets concentrated at its main channel.
Cluster 2 – Small watersheds with moderate to heavy soils. San Enrique, Suay, Tayuman, Malijao, Tinihaban, Talave, and Andoon are the watersheds belonging to this cluster. What makes this cluster distinct is their very high proportion of moderate and heavy-textured soils, which is almost 100% of their total catchment area. Clay, hydrosol, clay loam, loam, and rough mountainous land are their dominant soil textures. In terms of soil hydrologic group, they have Group B and D soils which have low to high runoff potential. The different combinations of the proportion of their clayey and loamy soils render these watersheds to very low to very high erodibility.
These are small watersheds with less than 100 km2 catchment area. Though these are quite scattered around the island, these are found to go by pairs, beside each other; Tayuman and Suay located at the south-western Table 5. Analysis of variance showing contribution of each component factor.
Factor Cluster Error F P-Value
Mean Square Df Mean Square Df
Drainage Shape Relief SizeGeology Stream Soils Land cover
2.963 2.876 1.033 3.179 2.796 1.783 2.514 3.002
88 88 88 88
0.346 0.375 0.989 0.274 0.401 0.739 0.495 0.333
2424 2424 2424 2424
8.577 7.675 1.044 11.608
6.968 2.412 5.078 9.021
0.000 0.000 0.432 0.000 0.000 0.046 0.001 0.000 Table 6. Final cluster centers highlighting the factors with higher contribution in each cluster.
Factor Regression Factor Score for Each Cluster
1 2 3 4 5 6 7 8 9
Drainage Shape Relief SizeGeology Stream Soils Land cover
1.824 -0.110 -1.087 3.693 0.430 0.061 0.058 0.577
0.156 0.089 -0.008 -0.682 0.502 -0.668
1.029 0.087
-0.468 -0.003 -1.482 -0.911 0.389 2.247 0.370 -0.196
-0.477 -1.586 -0.102 -0.035 -0.831 0.062 -0.077
1.322
0.314 -1.346 0.562 0.085 -1.850 -0.769 0.092 -3.850
-0.395 0.815 0.439 1.147 -0.454 0.269 0.842 -0.155
0.389 -0.597 -0.178 0.017 1.381 -0.108 -0.850 -0.449
1.158 1.111 0.169 -0.527 -0.840 0.093 -1.008
0.298
-1.634 0.340 0.560 0.228 -0.109 -0.090 -0.467 -0.214
Figure 4. The spatial distribution of the nine watershed clusters.
28
side, San Enrique at the west-central side, Malijao and Tinihaban at the north-western side, and Talave and Andoon at the eastern side.
Cluster 3- Watersheds with many but shorter lower order streams. Watersheds of this cluster have the highest recorded mean bifurcation ratio and stream length ratio values. Sum-ag and Vito watersheds have third order stream network with shorter but many lower order streams. This stream network characteristic is favorable to flood but because these have low relief (30- 70 mASL), this is cancelled out. High runoff may occur in the upstream areas where the first order streams meet the second order streams but because the topography is not steep at the downstream section and the receiving higher order stream is longer, a higher infiltration rate is expected. Despite being low relief catchments, Sum-ag watershed has less than 1%, while Vito watershed has no alluvium deposits.
Cluster 4- Highly-urbanized watersheds. This cluster of watersheds is distinct for it is greatly influenced by its location in highly-urbanized areas. Lupit, Magsungay, and a portion of Matab-ang watersheds drain the city proper of Bacolod City, the capital of the province. These have a relatively high percentage of built-up area and zero forest cover. Their high percentage of impervious land surface increases their runoff potential and the risk of flooding. Moreover, this cluster is also characterized with high percentage of light-textured soils. More than 80% of their soil cover has sandy-loam texture. This soil texture assures good drainage. In terms of soil factor, the watersheds have low flood susceptibility but their lightness in terms of structure makes them more susceptible to erosion.
These watersheds are small-sized watersheds with almost similar catchment shape, drain from the same recharge area, and are spatially adjacent to each other. But just a watershed away, Malogo watershed, a medium-sized catchment with a unique catchment shape, is located. It was included in this cluster despite its contrast with the other three watersheds, making it a group outlier. It could have been counted in this cluster because it showed unique characteristics in its land cover. Among the thirty-three watersheds, it is one of the eleven watersheds with remaining forest cover and the only watershed with more than 20% of its drainage area covered with forest. Its forest cover is around 55% of its drainage area, while its urban area is just around 3.45%.
Cluster 5- Watershed with high maximum elevation.
Marayo watershed is another stand-alone watershed
belonging to a cluster where watershed relief characteristics are distinctive. Marayo is a medium-sized watershed which recorded the highest ruggedness number and relief ratio. These two relief parameters are strongly influenced by the watershed’s maximum elevation. Its watershed shape is also distinctive which is likened to a tadpole. Its tail, indicating its upstream area, is long, narrow and steep. Its remotest end has an elevation of about 2,350 mASL located at Mt. Kanlaon and it slides down between two large watersheds, Bago and Binalbagan River basins. Despite its very high ruggedness number and relief ratio values or very steep upstream area, the watershed has low susceptibility to flood due to its very low hypsometric integral, indicating that more than 90% of its watershed area is located at low elevation.
Cluster 6- Large, elongated watersheds with medium to heavy soils. This cluster includes a river basin (Binalbagan), a large watershed (Bago), and two medium- sized watersheds (Himogaan and Himamaylan) with drainage areas occupying the northern to central spine of the island. The larger watersheds have a 5th order river system, while Himamaylan is a 4th order river system.
Aside from their distinctive large size, these also have moderate to heavy textured soils, mainly clayey, loamy and rough mountainous lands. The three larger watersheds have relatively high percentage of hydrologic soil group D indicating higher runoff potential, while Himamaylan watershed has higher group B soils which makes it less susceptible to flood. In terms of soil erodibility, the watersheds have very low to moderate susceptibility.
Their watershed shape parameters values also indicated that these are elongated catchments, except for Himamaylan watershed, which is less elongated in shape.
Cluster 7- Agricultural watersheds underlain with sedimentary rock. Five watersheds belong to this cluster, Danao, Calatrava, Tanao, Salamanca, and Paton-an.
These watersheds have small to medium catchment areas (30-150 km2) located very close to each other clumped together at the north-eastern side of the island. More than 90% of their catchment area is cultivated to crops, thus these are considered agricultural catchments. Except for the Tanao watershed, which could be considered as an outlier in the group, the four watersheds are predominantly underlain with limestone, shale and sandstone, which are common lithologic units found at the north-eastern and south-western sides of the island. Geologically, these watersheds have low susceptibility to flood and erosion.
Cluster 8- Elongated watersheds with coarse drainage network. This watershed cluster is distinct from other clusters due to its drainage characteristics
29 and catchment shape. In terms of shape, Sipalay, Isio,
Cawayan, Huyabhuyab, and Hinigaran watersheds have the highest elongation ratio values that range from 0.63 to 0.74. These values signify elongated to less elongated shape and a relatively higher susceptibility to flood and erosion. Their less elongated shape could be observed with their pear-shaped catchments. The influence of their catchment shape on susceptibility, however, is canceled out with their other geomorphological characteristics, such as their drainage characteristics. These recorded the lowest drainage density and stream frequency values, which signify fewer and shorter stream segments relative to their catchment area. With less number and farther apart channels, a coarse drainage pattern is produced. These also have the highest length of overland flow values which also signifies less contribution to runoff and erosion. Except for Hinigaran watershed, Huyabhuyab, Sipalay, Isio, and Cawayan watersheds are clustered together at the south-western side of the island.
Cluster 9- Medium-sized watersheds with high maximum elevation. Watersheds belonging to this cluster are defined by their catchment relief characteristics. These are medium-sized watersheds with high maximum elevations: Imbang (1,660 mASL), Sicaba (1,520 mASL), and Talabaan (1,330 mASL). Their high maximum elevation contributed to their high relief ratio values, very high ruggedness number values and moderate mean basin elevation values. Their upstream catchment begins at the mountain slopes of Mt. Silay and Mt. Mandalagan and their outlets drain at the northern coastline of the island. Considering the contribution of other geomorphologic factors, these catchments may be moderately susceptible to flood and erosion.
Tinampaan watershed is also a cluster member of this group despite its relatively lower maximum elevation (400 mASL). It is just a small-sized watershed with moderate relief ratio and ruggedness number values, and low mean basin elevation value. The closeness to other bigger watersheds that could be observed is only its spatial proximity to Talabaan and Sicaba watersheds.
Spatial Proximity and Physical Similarity
Spatial proximity and physical similarity are two separate approaches often used in regionalization studies in the selection of donor catchment for runoff predictions of ungauged catchments. Together with regionalization by regression as the third approach, these three approaches are compared as to which could closely predict the hydrologic behavior of the ungauged watersheds. Oudin et al. (2008) found that spatial proximity approach is better than
physical similarity and that regression approach is least satisfactory. This result was confirmed in the studies of Merz and Blöschl (2005); Zhang and Chiew (2009);
Bao et al. (2012); and Drogue et al. (2016). These also suggested that an integrated or combined spatial proximity and physical similarity could help improve the estimation.
This classification is mainly based on the physical similarity approach, where watersheds were grouped according to their most similar attributes. However, when shown on a spatial map, some of the watersheds of the same group seem to be closer to each other. Thus, it is interesting to know whether watersheds with similar attributes that are grouped together in a cluster are also spatially close to each other in the case of an island region.
The distance of the watersheds from their cluster centers based on the cluster membership data (which represents physical similarity) was compared with the measured mean spatial distance of the watersheds relative to other cluster members, (which represents spatial proximity).
Pearson’s correlation analysis showed that at 0.01 level of significance, there is a moderate positive correlation between the two variables (Pearson’s r= 0.509 at p=
0.002) indicating that most watersheds classified in clusters based on their strong physical similarity are also spatially close to each other.
Watersheds considered as outliers can be identified and these had substantial effect on the fit of the curve.
(Figure 5) The scatter plots show the dispersion of the points representing the relative distances of the watersheds from one another. Outliers were omitted and an improved correlation value at p = 0.0005 of Pearson’s r = 0.841 appeared, indicating a strong correlation between the two compared factors.
Correlation analysis was also conducted for each cluster with more than two member watersheds (Table 7). Cluster 4, the highly urbanized watersheds with the forested Malogo watershed, got the highest correlation coefficient of r= 0.99. This is followed by Cluster 8 composed of elongated watersheds with coarse drainage pattern (r = 0.85), Cluster 6 of large, elongated watersheds with medium to heavy soils (r = 0.84), and Cluster 2 composed of small watersheds with moderate to heavy soils (r = 0.79). Cluster 9, medium-sized watersheds with high maximum elevation, and Cluster 7, agricultural watersheds, got the lowest coefficient of correlation values of r = 0.45 and r = 0.36, respectively.
Despite the strong correlation relationships discovered in most watershed clusters, only Cluster 4 has
30
CONCLUSIONS AND RECOMMENDATIONS Classifying watersheds into groups with similar catchment biophysical characteristics does not only promote simplicity but also facilitates and provides explanation and understanding of the nature of not only one watershed but also its relationship with other watersheds in a bigger landscape. Nine watershed clusters were established, described, and mapped out with watershed size having the greatest contribution to the grouping and followed by land cover, drainage characteristics, watershed shape, catchment geology, and soil variables.
Analysis showed a positive relationship between spatial proximity and physical similarity of watersheds in a given region. This confirmed the general observation that regional catchments are relatively homogenous and share similar characteristics and this holds true too in an island region.
The main contribution of this study for the province is the establishment of a comprehensive database for its major watersheds. The characterization has covered important information that could be utilized for island- wide planning and management of water and soil resources using watershed as a management unit. At present, only a few of these watersheds were characterized and for different purposes and contexts. But putting all these major watersheds in one picture together in spatial thematic maps, the government, community, and academe in the province would have something concrete to work with. It is hoped that with these information, Figure 5. Scatterplots of clustered watersheds showing relationship between physical similarity and
spatial proximity with (left) and without outliers (right).
Table 7. Correlation values computed for clusters with more than two watersheds.
Cluster Pearson
Coefficient p-Value N 24
67 89
0.702 0.990*
0.841 0.363 0.849 0.418
0.120 0.010 0.364 0.549 0.069 0.582
64 35 54
strong significant correlation with p value of 0.01. The high estimate of correlation coefficient values with high p-values means high correlation with low probability.
This could be attributed to the small sample size (N) for each cluster.
Generally, the values show that most of the watersheds that were clustered not only share similar features but are also near each other. This could also mean that these watersheds share some similar physical features because these are located at the same region and were formed and are continually affected by the same processes of nature. This is in line with the hypothesis that regional catchments are relatively homogenous and have similar characteristics (Bao et al. 2012). Hence, the use of physical similarity approach that also take into consideration the closeness of watersheds based on their characteristics is a reasonable approach not only in regionalization approach (Heřmanovskỳ and Pech 2008) but also in watershed classification.
31 Bayrd, G.B. 2006. The Influences of Geology and Water
Management on Hydrology and Fluvial Geomorphology in the Henry’s Fork of the Snake River, Eastern Idaho and Western Wyoming. Unpublished Graduate Thesis.
Idaho State University.
Beven, K.J. 2000. “On the Uniqueness of Place and Process Representations in Hydrological Modeling”. Hydrology and Earth System Science 42: 203-212.
Breinlinger, R., Düster, R. and Weingartner, R. 1996.
“Methods of Catchment Characterization by Means of Basin Parameters Assisted by GIS – Empirical Report from Switzerland”. In: Methods of Hydrological Basin Comparison, Report No. 120. (ed. M. Robinson). Paper presented at the Fourth Conference of the European Network of Experimental and Representative Basins.
University of Oxford. September 1996.
Brocklehurst, S.H. and Whipple, K.X.. 2004. “Hypsometry of Glaciated Landscape”. Earth Surface Processes and Landforms 29: 907-926.
Brown, S.C., Lester, R.E., Versace, V.L., Fawcett, J.
and Laurenson, L. 2014. “Hydrologi Landscape Regionalization Using Deductive Classification and Random Forests”. PLoS ONE 9(11): 1-20.
Bureau of Soils and Water Management (BSWM). 2005. Soil Survey Report: Negros Occidental Province. Soil Survey Division, BSWM, Department of Agriculture.
Carillo, G., Troch, P.A., Sivapalan, M., Wagener, T., Harman, C. And Sawicz, K. 2011. “Catchment Classification:
Hydrological Analysis of Catchment Behavior through Process-based Modeling along a Climate Gradient”.
Hydrology and Earth System Sciences 15: 3411-3430.
Chess, C. and Gibson, G. 2001. “Watersheds are not Equal:
Exploring the Feasibility of Watershed Management - Reply”. Journal of the American Water Resources Association 37, 775 - 782.
Chorley, R. J., Donald, E.G. and Pogorzelski, H.A. 1957. “A New Standard for Estimating Drainage Basin Shape”.
American Journal of Science 255: 138-141.
Choubin, B., Solaimani, K., Habibnejad, R.M., Malekian, A.
2017. “Watershed Classification by Remote Sensing Indices: A Fuzzy C-Means Clustering Approach”.
Journal of Mountain Science 14(10): 2053-2063.
Davis, J. 2010. “Hypsometric Tools Software”. Downloaded from arcscripys.esri.com/ details.asp?dbid=16830 Department of Environment and Natural Resources (DENR).
2008. Guidelines in the Preparation of Integrated Watershed Management Plans, DENR Memorandum the provincial government would: include watershed
management in its priority list of programs covering both small and large watersheds at risk, making watershed as their unit for planning and decision-making; conduct watershed studies to further identify common issues and threats on water, soil, and people, and assess condition of clustered watersheds; establish co-management agreements among cities and municipalities drained by the same watershed and provide guidance and mechanisms in the management process; identify other available natural and social resources, socio-political and cultural structures, networks and dynamics, and existing social and environmental issues and threats in each watershed to improve the established database; and increase awareness and ownership of the community through information, education and communication campaign about the characteristics and ecological functions of their watershed in a form best understood and appreciated by the local community.
Results of catchment classification could be further improved if other parameters were considered to further understand how catchment structures and external factors could affect the catchment hydrologic functions.
The limitation of the study presented here is the focus on the catchment biophysical characteristics only due to the limited availability of hydrologic and climate data and more recent datasets from government offices. However, the simple clustering methodology could serve as a guide to other islands or regions as a first step to decrease the complexity of the catchments given limited available data.
REFERENCES
Ali, G., Tetzlaff, D., Soulsby, C., McDonnell, J.J. and Capell, R. 2012. “A Comparison of Similarity Indices for Catchment Classification Using a Cross-Regional Dataset”. Advances in Water Resources 40: 11-22.
Al-Rawas, G. and Valeo, C. 2010. “Relationship between Wadi Drainage Characteristics and Peak Flood Flow in Arid Northern Oman”. Hydrological Sciences Journal 553:
377-393.
Bao, Z., Zhang, J., Liu J., Fu, G., Wang, G., He, R., Yan, X., Jin, J., Liu, H. 2012. “Comparison of Regionalization Approaches Based on Regression and Similarity for Predictions in Ungauged Catchments under Multiple Hydro-climatic Conditions”. Journal of Hydrology 466–
467: 37–46.
Bao, T.Q. and Laituri, M.J. 2013. “The Effects of Watershed Characteristics on Storm Runoff Relationships in Vietnam”. Journal of Environmental Science & Water Resources 22: 040-052.
32
Circular Number 2008-05, Diliman, Quezon City.
Drouge, G., Ben-Khediri, W., and Conan, C. 2016. “Added- Value from a Multi-Criteria Selection of Donor Catchments in the Prediction of Continuous Streamflow Series at Ungauged Pollution Control Sites”. Paper presented at the 7th International Water Resources Management Conference of ICWRS. Bochum, Germany.
18–20 May 2016.
Guadalquiver, N. and Nicavera, E. 2019. NegOcc provincial government, business group back sugar industry.
Philippine News Agency. https://www.pna.gov.ph/
articles/1083350
Hidayatulloh, A., Bamufleh, S., Chaabani, A., Al-Wagdany, A., and Elfeki, A. 2021. “Affinity Propagation Approach for Catchment Classification Applied to Arid Catchments”.
Journal of African Earth Sciences 184: 104374.
Horton, R.E. 1945. “Erosional Development of Streams and their Drainage Basin: Approach to Quantitative Morphology”. Geological Society of America Bulletin 56: 275-320.
Horton, R.E. 1932. “Drainage Basin Characteristics”.
Transactions, American Geophysical Union 13: 350- 361.
Mayer, A., Winkler, R. and Fry, L. 2014. “Classification of Watersheds into Integrated Social and Biophysical Indicators with Clustering Analysis”. Ecological Indicators 45: 340-349.
Mazvimavi, D. 2003. Estimation of Flow Characteristics of Ungauged Catchments: Case Study in Zimbabwe.
Unpublished Ph.D. Dissertation. Wageningen University, The Netherlands.
McDonnell, J.J. and Woods, R. 2004. “On the Need for Catchment Classification”. Journal of Hydrology 299: 2-3.
Melton M.A. 1958. “Correlation Structure of Morphometric Properties of Drainage System and their Controlling Agents”. Journal of Geology 66: 442-460.
Merz, R. and Blöschl, G. 2005. “Flood Frequency Regionalization - Spatial Proximity vs. Catchment Attributes”. Journal of Hydrology 302(1-4): 283-306.
Miller, V.C. 1953. A Quantitative Geomorphic Study of Drainage Basin Characteristics in the Clinch Mountain Area, Virginia and Tennessee. Technical Report No. 3.
Columbia University. Department of Geology, ONR, Geography Branch, New York.
Mines and Geosciences Bureau (MGB). 1981. Geology and Mineral Resources of the Philippines 1st Ed. Ministry of
Natural Resources, Manila, Philippines.
O’Neill, K.M. 2005. “Can Watershed Management Unite Town and Country?” Society and Natural Resources 183: 241-253.
Oudin, N, L., Andreassian, V., Perrin, C., Michel, C., Le Moine, N. 2008. “Spatial Proximity, Physical Similarity, Regression and Ungaged Catchments: A Comparison of Regionalization Approaches Based on 913 French Catchments”. Water Resources Research 44:3.
Raux, J., Copard, Y., Laignel, B., Fournier, M., Massei, N. 2011.
“Classification of Worldwide Drainage Basins through the Multivariate Analysis of Variables Controlling their Hydro-Sedimentary Response”. Global Planetary Change 76: 117-127.
Razavi. T. and Coulibaly, P. 2013. “Classification of Ontario Watersheds Based on Physical Attributes and Streamflow Series”. Journal of Hydrology 493: 1-94.
Salami, Y. And Buehler, A. 2020. “Catchment Classification in a Transboundary River Using Run-off and Sub-basin Characteristics”. International Journal of River Basin Management 18(1): 117-126.
Sawicz, K., Wagener, T., Sivapalan, M., Troch, P.A., and Carillo, G. 2011. “Catchment Classification: Empirical Analysis of Hydrologic Similarity Based on Catchment Function in the Eastern USA”. Hydrology and Earth System Sciences 15, 2895-2911.
Sawicz, K.A. 2013. Catchment Classification- Understanding Hydrologic Similarity Through Catchment Function.
Published Ph.D. Dissertation. The Pennsylvania State University, USA. https://www.proquest.
com/dissertations-theses/catchment-classification- understanding-hydrologic/docview/1467508727/se- 2?accountid=190473
Schumm, S.A. 1956. “Evolution of Drainage Systems and Slopes in Badlands at Perth Amboy, New Jersey”.
Geological Society of America Bulletin 67: 597-646.
Sivakumar, B., Singh, V., Berndtsson, R. and Khan, S. 2014.
“Catchment Classification Framework in Hydrology:
Challenges and Directions”. Journal of Hydrologic Engineering 20: A4014002-1-12.
Strahler, A.N. 1952. “Dynamic Basis of Geomorphology”.
Geological Society of America Bulletin 63: 923-938.
Strahler, A.N. 1964. “Quantitative Geomorphology of Drainage Basins and Channel Networks, Section 4II”.
In: Handbook of Applied Hydrology. (ed. V.T. Chow).
McGraw Hill. pp. 439.
33 Strahler, A.N. 1968. “Quantitative Geomorphology”.
In: Fairbridge, R.W., Ed., The Encyclopedia of Geomorphology, Reinhold Book Crop., New York.
Van Hoek, M. 2019. “Understanding Classification of European Catchments using Artificial Intelligence”. Geophysical Research Abstracts 21, 1-1.
Vasquez, N., Cepeda, J., Gomez, T., Mendoza, P., Alvarez, C., Lagos, M., and Vargas, X. 2019. “Exploring the Relation Between Meteorological, Physiographic and Hydrological Similarities through Catchment Classification”. Geophysical Research Abstracts 21, 1.
Wagener, T., Sivapalan, M.,Troch, P., Woods, R. 2007.
“Catchment Classification and Hydrologic Similarity’.
Geography Compass 14: 901-931.
Yi, J., Jintao, L., Lin, L., Aihua, W., and Xi, C. 2018.
“Exploring Hydrologically Similar Catchments in Terms of the Physical Characteristics of Upstream Regions”.
Hydrology Research 49(5): 1467-1483.
Zhang, Y. and Chiew, F.H.S. 2009. “Relative Merits of Different Methods for Runoff Predictions in Ungauged Catchments”. Water Resources Research 45: 1-13.
ACKNOWLEDGMENT
This study was made possible through the assistance of Accelerated Science and Technology Human Resource Development Program (ASTHRDP) scholarship program of Department of Science and Technology – Science Education Institute (DOST-SEI), University Consortium Research Grant of Southeast Asian Regional Center for Graduate Study and Research in Agriculture (SEARCA), and University Research Center of University of St. La Salle-Bacolod.
